HMGA1 Amplifies Wnt Signalling and Expands the Intestinal Stem Cell

ARTICLE

Received 8 Dec 2016 | Accepted 21 Feb 2017 | Published 28 Apr 2017 DOI: 10.1038/ncomms15008 OPEN HMGA1 ampliﬁes Wnt signalling and expands the intestinal stem cell compartment and Paneth cell niche

Lingling Xian1, Dan Georgess2, Tait Huso1, Leslie Cope3, Amy Belton1, Yu-Ting Chang4, Wenyong Kuang1, Qihua Gu1, Xiaoyan Zhang1, Stefania Senger5, Alessio Fasano5, David L. Huso6,z, Andrew J. Ewald2,7 & Linda M.S. Resar1,7,8

High-mobility group A1 (Hmga1) chromatin remodelling proteins are enriched in intestinal stem cells (ISCs), although their function in this setting was unknown. Prior studies showed that Hmga1 drives hyperproliferation, aberrant crypt formation and polyposis in transgenic mice. Here we demonstrate that Hmga1 ampliﬁes Wnt/b-catenin signalling to enhance self- renewal and expand the ISC compartment. Hmga1 upregulates genes encoding both Wnt agonist receptors and downstream Wnt effectors. Hmga1 also helps to ‘build’ an ISC niche by expanding the Paneth cell compartment and directly inducing Sox9, which is required for Paneth cell differentiation. In human intestine, HMGA1 and SOX9 are positively correlated, and both become upregulated in colorectal cancer. Our results deﬁne a unique role for Hmga1 in intestinal homeostasis by maintaining the stem cell pool and fostering terminal differentiation to establish an epithelial stem cell niche. This work also suggests that deregulated Hmga1 perturbs this equilibrium during intestinal carcinogenesis.

1 Division of Hematology, Department of Medicine, The Johns Hopkins University School of Medicine, 720 Rutland Avenue, Ross Research Building, Room 1025, Baltimore, Maryland 21205, USA. 2 Department of Cell Biology, The Johns Hopkins University School of Medicine, 855 North Wolfe Street, Baltimore, Maryland 21205, USA. 3 Division of Biostatistics, Department of Oncology, The Johns Hopkins University School of Medicine, 550 North Broadway, Baltimore, Maryland 21205, USA. 4 Department of Pathology, Pathobiology Graduate Program, The Johns Hopkins University School of Medicine, 720 Rutland Avenue, Ross Research Building, Room 1025, Baltimore, Maryland 21205, USA. 5 Department of Pediatrics, Mucosal Immunology and Biology Research Center, Harvard Medical School, Massachusetts General Hospital East, 16th Street, Building 114, Charlestown, Massachusetts 02114, USA. 6 Department of Molecular and Comparative Pathobiology, The Johns Hopkins University School of Medicine, Baltimore, Maryland 21205, USA. 7 Department of Oncology, The Johns Hopkins University School of Medicine, Baltimore, Maryland 21205, USA. 8 Department of Pathology and Institute for Cellular Engineering, The Johns Hopkins University School of Medicine, Baltimore, Maryland 21205, USA. Correspondence and requests for materials should be addressed to L.M.S.R. (email: [email protected]). zDeceased.

NATURE COMMUNICATIONS | 8:15008 | DOI: 10.1038/ncomms15008 | www.nature.com/naturecommunications 1 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms15008

ntestinal stem cells (ISCs) provide a paradigm for studying We discover that Hmga1 expands the ISC pool and Paneth cell adult stem cell function due to their exceptional self-renewal niche in vivo. Hmga1 is a key factor involved in the organization Ipotential and repetitive structural organization1–5. Indeed, the of ISCs into three-dimensional (3D) organoids in vitro. We also intestinal lining is among the most highly regenerative tissues, find that Hmga1 enhances ISC expansion and self-renewal by renewing itself every 3–5 days to protect the gut from pathogens amplifying Wnt/b-catenin signalling. Hmga1 also directly and maintain nutrient intake essential for life. Over the past upregulates Sox9 and expands the Paneth cell niche. This is an decade, a population of self-renewing, columnar epithelial cells example of Hmga1 fostering terminal differentiation to establish a located at the base of the intestinal crypts has been identified and stem cell niche. Moreover, both HMGA1 and SOX9 are positively characterized as ISCs1–5. They are marked by the serpentine correlated in human intestinal epithelium, and both become receptor, leucine-rich repeat containing G-protein-coupled markedly upregulated in colorectal cancer. These results reveal a receptor 5 (Lgr5), which mediates Wnt signalling cues from the unique role for Hmga1 in maintaining both the ISC pool and niche5. Lineage tracing experiments demonstrate that these ISCs niche cells within intestinal crypts and suggest that this are responsible for the exuberant regeneration and tissue equilibrium is perturbed when Hmga1 becomes deregulated homeostasis in intestinal epithelium1,4,6. Despite extensive during carcinogenesis. study, the molecular mechanisms that govern their behaviour are only beginning to be elucidated1–9. Previous work also demonstrates that aberrant expression or mutation of key Results regulators of ISCs leads to neoplastic growth and intestinal Hmga1 drives expansion of the ISC compartment. A prior gene carcinogenesis10,11. expression profile study showed that Hmga1 is among the genes Emerging evidence highlights the central role for chromatin enriched in Lgr5 þ ISCs (ref. 29). HMGA1 is also among the genes structure and chromatin-binding proteins in maintaining stem most highly expressed in diverse epithelial human cancers cell properties. In fact, recent work found that the high-mobility as compared to normal epithelium, including intestinal group A1 chromatin remodelling proteins (HMGA1, formerly malignancies12–14,17,33. We therefore sought to elucidate the HMG-I/Y) regulate stem cell properties in cancer12–18, although functional role of Hmga1 in ISCs, both in normal intestinal their role in normal development has remained elusive. The epithelial homeostasis and in intestinal neoplasia. To this end, HMGA1 gene encodes the HMGA1a and HMGA1b isoforms19–21, we crossed our Hmga1 transgenic mice onto Lgr5-EGFP mice6, which function as architectural transcription factors. HMGA1 which mark Lgr5 þ ISCs with enhanced green fluorescent protein proteins bind to specific DNA sequences13,22–24, modulate (EGFP). The Hmga1 transgene is driven by the H-2Kb promoter chromatin structure23 and recruit other transcriptional and m enhancer, which confer transgene expression in intestinal complexes to regulatory regions throughout the genome13,22,23. crypt basilar cells37,lymphoidcells35 and uterine tissue36.Inboth HMGA1 is highly expressed during embryogenesis, with high Hmga1 transgenic and wild-type (WT) mice, Hmga1 protein levels in normal embryonic stem cells13,16,25,26. Postnatally, localizes to the nuclei of Lgr5 þ ISCs (Fig. 1a–d). Interestingly, HMGA1 is expressed in adult stem cells, such as hemato- Lgr5 þ ISCs extend further up the crypts in the Hmga1 transgenic poietic27,28 and intestinal stem cells29, but absent or barely mice compared to WT mice in all regions of the small intestine detectable in mature, differentiated tissues. In cancer, HMGA1 (duodenum, jejunum, ileum; Fig. 1a–f), consistent with the becomes aberrantly expressed through oncogenic transcription expansion in the ISC pool and enhanced self-renewal in vivo. factors and epigenetic alterations, or in rare cases, chromosomal The percentage of Lgr5 þ ISCs per total crypt cells by fluorescent translocation events13,17,30,31. Moreover, HMGA1 is over- stain (Fig. 1f; **Po0.00001; Mann-Whitney test) and relative expressed in most high-grade or poorly differentiated cancers frequency of Lgr5 þ ISCs per crypt cell isolates by fluorescence- studied to date, and high levels portend a poor prognosis in activated cell sorting (FACS) were significantly (Po0.05; two-tailed diverse tumours12–18,26,31–36. In murine tumour xenografts, Student’s t-test) increased in the Hmga1 transgenic small intestine HMGA1 drives tumour progression and cancer stem cell (Fig. 1g,h), further validating the expansion in ISCs. To determine properties, at least in part, by inducing stem cell transcriptional whether Hmga1 gene expression accounts for the elevated protein networks12–18. In human embryonic stem cells, HMGA1 levels in the Lgr5 þ ISCs, we assessed Hmga1 mRNA in WT mice maintains a de-differentiated state by upregulating genes and found that it is increased by B4-fold in Lgr5 þ ISCs isolated involved in stemness and pluripotency16. Moreover, HMGA1 is by FACS compared to Lgr5 À cells (Fig. 1g,h). We also assessed required for reprogramming somatic cells to induced pluripotent Hmga1 mRNA in crypt cells from the transgenic model and found stem cells by the Yamanaka factors; disrupting HMGA1 that Hmga1 expression is increased about twofold in both Lgr5 þ expression or function prevents the derivation of fully ISCs and Lgr5 À cells compared to the same cell populations in reprogrammed cells16. Given its dual role in normal WT mice. In both WT and transgenic models, Hmga1 mRNA is development and cancer, further studies to dissect HMGA1 enriched in the Lgr5 þ ISC population (Fig. 1g,h). function in each setting are needed to determine the therapeutic ISCs are regulated by factors from the stromal compartment in potential of targeting HMGA1 in cancer or harnessing its function addition to intestinal epithelial cells1–9,38–40. To define the role of for tissue regeneration. Hmga1 in ISCs within the epithelial compartment, we used We previously demonstrated that transgenic mice overexpres- organoids, an in vitro intestinal crypt cell culture model1–4,41. sing murine Hmga1 from the H-2Kb promoter and immunoglo- Organoid buds are a surrogate for ISC function because they bulin m enhancer all succumb to lymphoid tumours35; females comprise crypt-like structures with ISCs on the tips; differentiated also develop uterine sarcomas36. In this model, the transgene is epithelial cells extend towards the luminal centres of the expressed in the intestines14 in addition to lymphoid cells35 and organoids. We derived organoids from small intestinal epithelial uterine tissue36. The Hmga1 transgenics develop marked crypt cells isolated from transgenic or WT mice and compared proliferative changes in the epithelium of the small and large bud formation and projected surface area. Prior to culture, an intestine, with aberrant crypt formation and polyposis14. equal number of crypts from WT or Hmga1 transgenic mice of To determine how Hmga1 disrupts tissue homeostasis in similar sizes were isolated by passage through a 70 mM filter41. the intestines of transgenic mice and intestinal cancers Similar to the histologic results in the transgenic mouse intestine, overexpressing HMGA1, we examined its expression and we found a marked increase in projected surface area per function in our transgenic model and in intestinal organoids. organoid (Image-Pro Plus Version 6) and bud number per

2 NATURE COMMUNICATIONS | 8:15008 | DOI: 10.1038/ncomms15008 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/ncomms15008 ARTICLE

WT/Lgr5-EGFP(+) Hmga1/Lgr5-EGFP(+)

acGFP b Hmga1 GFP d Hmga1

Hmga1/GFP Hmga1 Hmga1/GFP Hmga1

e f Duodenum (Duo) Jejunum (JE) lleum (IL) 100 GFP/DAPI 80

60 WT 40

20 GFP/DAPI WT 80 Hmga1 70

Hmga1 60 ** ** **

% GFP(+) cells per crypt 50 40 gh30 20 * 10 12 * 0 Duo JE lL WT 10 SSC-A * 3.0% 8 ± 0.08% * mRNA 6

4 Hmga1 2 SSC-A Hmga1 5.2% 0 Hmga1 ± 0.2% WT GFP G(–) G(+) G(–) G(+)

Figure 1 | Hmga1 localizes to ISCs and its overexpression expands the ISC compartment in transgenic mice. (a) ISCs from WT/Lgr5-EGFP þ mice stain green (upper panel) in small intestinal cross-sections. Hmga1 stains red and localizes to GFP þ ISCs (lower panel). (b) Intranuclear Hmga1 stains brown (immunohistochemistry (IHC)) in small intestinal cross-sections from WT/Lgr5-EGFP þ mice. (c) ISCs stain green in Hmga1/Lgr5-EGFP þ transgenic mice and extend further up the base of the crypt in Hmga1 transgenic mice compared to WT mice. (Cells with red nuclear staining outside of crypts are Hmga1 transgenic lymphocytes) (d) Intranuclear Hmga1 stains brown (IHC) in small intestinal cross-sections from Hmga1/Lgr5-EGFP þ mice. (e) Fluorescent staining identifies GFP þ ISCs at each region in the small intestine of WTand Hmga1 mice (duodenum, jejunum, ileum). (f) The percentage (mean±s.d.) of GFP þ ISCs per crypt in WT and Hmga1 transgenic mice are shown; **Po0.00001, Mann–Whitney test (n ¼ 50 crypts per region, 3 mice per genotype). Dot plot shows individual data points. (g) Hmga1 transgenic mice have higher GFP þ ISC frequency as assessed by flow cytometry; mean frequency±s.d. from two experiments are shown. (h) Hmga1 mRNA is enriched in ISCs isolated by flow cytometry for GFP þ cells using quantitative, real-time PCR (qPCR) in WT and Hmga1 transgenic mice. Hmga1 is higher in both GFP þ ISCs and GFP À crypt cells from the Hmga1 transgenic model compared to WT mice. Bars show mean relative Hmga1 expression±s.d. from three experiments performed in triplicate; Gapdh was used as a loading control; *Po0.05; two-tailed Student’s t-test. Scale bars, 20 mm. organoid in those derived from the Hmga1 transgenic mice as performed time-lapsed confocal microscopy42–45 to ascertain the compared to WT controls (Fig. 2a–c; Supplementary Fig. 1a,b). rate of self-renewal in Lgr5 þ ISCs and bud development from There were more Hmga1 organoids with 43 buds as compared day 0 to day 6 in 3D organoid culture (Fig. 2d–f). A representative to WT organoids by day 6 in culture (Fig. 2b). The majority of image of crypt cells from days 0 and 6 from WT mice or Hmga1 WT organoids had r1 buds per organoid, while the majority of transgenics, respectively, are shown (Fig. 2d). We discovered Hmga1 organoids had Z2 buds at day 6 (Fig. 2b). Next, we a marked increase in the rate of self-renewal at day 6 in the isolated small crypts by passage through a 40 mM filter40,41 and Lgr5 þ ISCs from the Hmga1 transgenics compared to the WT

NATURE COMMUNICATIONS | 8:15008 | DOI: 10.1038/ncomms15008 | www.nature.com/naturecommunications 3 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms15008

a bc ≥ = 2 12.0 WT Hmga1 3 )

120 2 = 1 = 0 m **

* μ

100 ( 4 80 8.0 60 40 4.0 20 Projected surface area Projected surface per organoid ×10 % Organoids with buds 0 0 WT Hmga1 WT Hmga1 WT Hmga1 Day 2 Day 6

d WT Hmga1 e Day 0 Day 6 Day 0 Day 6 Merge 14 12 10 F-actin ** 8 ** 6 Lgr5-GFP 4 2

Nuclei ISCs of Lgr5+ Expansion rate 0 Day 0 606

WT Hmga1

f ) 2

m 6 μ Hmga1 ( 3

4 * WT

Lgr5+ projected surface Lgr5+ 0 area per organoid ×10 g 01234 Day 3 Day 8 Day 11 Bud number (Day 6) hi 0.12 ** **

WT 1.5 0.1

0.08 1 0.06 0.04 0.5 efficiency (%)

Colony forming Colony 0.02 Hmga1 0 0 WT Hmga1 Re-plating efficiency (%) WT Hmga1

Figure 2 | Hmga1 enhances ISC function in gut organoid cultures. (a) Typical 3D organoids cultured from crypts are shown from WTor Hmga1 transgenic mice. (b) Bud numbers were ascertained in 3D organoids from isolated crypts (n4100 organoids per mouse; 3 mice per genotype). Bars show mean percentage of organoids±s.d. with different bud numbers. (YellowZ3, grey ¼ 2, pink ¼ 1, blue ¼ 0). The mean percentage of organoids with Z 3 buds was increased in Hmga1 organoids compared to WT organoids. *Po0.05; two-tailed Student’s t-test. (c) Organoid projected surface area (PSA) is shown (mean±s.d.) from WT or Hmga1 mice (n420 organoids/genotype). **Po0.01; two-tailed Student’s t-test. (d) Representative confocal imaging of crypt cells (day 0) and organoids (day 6) with GFP þ staining for Lgr5 þ ISCs, Phalloidin (F-actin; red) for cell clusters and organoid structure, and Hoechst (blue) for nuclei are shown. (e) Relative expansion rate of ISCs was calculated by the ratio of the PSA of GFP þ ISCs on day 6 (nZ42 organoids per group) over the PSA on day 0 (nZ72 crypt cell clusters/group). **Po0.01; Mann–Whitney test. (f) ISC PSA (y axis) and bud number (x axis) at days 0 and 6 are shown. **Po0.01,*Po0.05; Mann–Whitney test. (g) Lgr5-GFP þ ISCs were purified by flow cytometry and cultured in matrigel. (h) Colony (organoid)-forming efficiency (mean±s.d.) was calculated from purified GFP þ ISCs isolated as single cells (n ¼ 3,000 cells per group) from WT and Hmga1 mice in three experiments; **Po0.01; two-tailed Student’s t-test. (i) Re-plating efficiency (mean±s.d.; n ¼ 3,000 cells per group) was assessed after dissociating colonies from purified GFP þ ISCs isolated as single cells from WTand Hmga1 mice using flow cytometry. **Po0.01; two-tailed Student’s t-test. Scale bars, 50 mm. mice based on the projected surface area of Lgr5 þ cells (mManager Previous work from our group demonstrated that forced software; Stanford Photonics; Fig. 2e,f). Similar to our prior results expression of Hmga1 prevents differentiation in human embryo- using standard microscopy, we also found more buds and greater nic stem cells16. We therefore reasoned that Hmga1 expression projected surface area of Lgr5 þ cells in the Hmga1 organoids would predominate at the crypt-like regions enriched for ISCs at versus the WT organoids by confocal microscopy (Fig. 2f). the bud tips and decrease in regions of differentiated cells at the

4 NATURE COMMUNICATIONS | 8:15008 | DOI: 10.1038/ncomms15008 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/ncomms15008 ARTICLE base of the buds. To test this, we assessed Hmga1 protein levels vector was induced by doxycycline express RFP and exhibit an throughout the organoids and found that Hmga1 is enriched at impaired capability to proliferate and generate 3D organoids; the the bud tips where ISCs predominate, and undetectable in the projected organoid surface area was decreased compared to differentiated cells, suggesting that differentiation is permitted in uninduced controls (Fig. 3d,e). Bud formation was also disrupted cells with lower levels of Hmga1 (Supplementary Fig. 2). (Supplementary Fig. 5b). Hmga1 gene silencing in organoids Together, these results suggest that Hmga1 is a key factor for transduced with the inducible shHmga1 was confirmed by ISC maintenance and function. quantitative, reverse transcription PCR (qPCR; Fig. 3e). To Because we could not exclude the possibility that the enhanced further validate these findings, we transduced organoids derived stem cell function observed in the Hmga1 organoids resulted from the Hmga1 mice with the shHmga1 vector and found that from in vivo exposure to lymphoid or other cells with transgenic silencing Hmga1 also disrupted their ability to form organoids Hmga1 expression and downstream factors, we engineered WT and generate buds (Supplementary Fig. 5c). These studies indicate organoids to overexpress Hmga1. To this end, we transduced that Hmga1 is a key factor involved in self-renewal, bud organoids via lentivirus expressing Hmga1 and GFP (FUGW- formation and organization into 3D organoid structures. Hmga1) and compared this to control organoids transduced with lentivirus expressing GFP alone (FUGW). The WT organoids engineered to overexpress Hmga1 exhibited a similar phenotype Hmga1 induces self-renewal through Wnt/b-catenin signalling. to the organoids derived from the Hmga1 mice with increased When organoids from Hmga1 mice were incubated with Wnt3a bud formation (Supplementary Fig. 3), demonstrating that this to enhance lentiviral transduction, we observed a striking phe- phenotype was dependent on Hmga1. notype whereby the Hmga1 organoids formed very large, cyst- To determine whether the enhanced stem cell function in like, spherical organoids comprised predominantly of Lgr5 þ organoid-forming ability and bud development resulted from ISCs, in contrast to WT organoids, which generated smaller cysts cell-autonomous properties of the Hmga1 ISCs or from ISC at a decreased frequency (Fig. 4a; Supplementary Fig. 6). These interactions with other crypt epithelial cells, we purified Lgr5 þ findings suggested that the Hmga1 organoids are more sensitive ISCs isolated as single cells from transgenic and WT mice using to Wnt signalling. Wnt/Tcf4/b-catenin signalling is an evolutio- flow cytometry and compared their efficiency in organoid narily conserved pathway important for self-renewal in epithelial formation and re-plating assays. The purified Hmga1 ISCs crypt ISCs and many other tissue-specific adult stem cells38. generated organoids at a fivefold greater efficiency as compared Moreover, hyperactive Wnt signalling leads to malignant to WT ISCs (Fig. 2g,h). Moreover, re-plating efficiency was also transformation in intestinal epithelium38. To begin to define the enhanced fivefold in the Hmga1 ISCs (Fig. 2i). Interestingly, the role of Hmga1 in Wnt/Tcf4/b-catenin signalling, we assessed Hmga1 organoids generated from purified Lgr5 þ ISCs adopt a immunostaining for b-catenin as a surrogate for canonical Wnt cyst-like, spherical structure similar to the morphology observed signalling. b-catenin was markedly increased in the Hmga1 in organoids treated with Wnt or engineered to overexpress Ascl2, transgenic intestinal epithelium and concentrated at the base of a transcription factor important in maintaining stem cell identity the crypts (Fig. 4b). A similar increase was also observed in WT in ISCs7. To determine whether Hmga1 organoids remained in a organoids transduced to overexpress Hmga1 (Fig. 4c). Since cyst-like structure because they were comprised predominantly of Hmga1 functions as an architectural transcription factor that undifferentiated Lgr5 þ stem cells, we stained for the ISC marker, alters gene expression, we hypothesized that Hmga1 could Lgr5-GFP þ . Strikingly, most of the cells within the Hmga1 upregulate expression of factors that enhance Wnt signalling. organoids generated from purified Hmga1 Lgr5-GFP þ ISCs To test this, we first compared expression of genes encoding Wnt maintain this stem cell marker (Lgr5-GFP þ ; Supplementary agonist receptors that function in intestinal epithelium, including Fig. 4), in contrast to WT organoids, which only stain for Lgr5- Lgr5, Frizzled (Fzd)5, Fzd7, low-density lipoprotein receptor- GFP þ at the tips of the buds (Fig. 2d). Together, these studies related protein 5 (Lrp5) and Lrp6 in purified Lgr5 þ ISCs isolated demonstrate that Hmga1 maintains undifferentiated Lgr5 þ ISCs from WT or transgenic mice. We found an increase in expression and enhances self-renewal in this model. of all Wnt agonist receptors tested (Lgr5, Fzd5/7, Lrp5/6) in the Next, we sought to determine whether Hmga1 is required for Hmga1 ISCs by 42 to 4-fold compared to control ISCs (Fig. 4d). self-renewal and ISC function. First, we silenced Hmga1 in WT Together, these results indicate that Hmga1 amplifies Wnt crypt cells using lentiviral-mediated delivery of short hairpin signalling by upregulating genes encoding Wnt agonist RNA (shRNA) targeting Hmga1 (shHmga1) and compared this to receptors in ISCs, including the ISC marker, Lgr5, in our model. WT crypt cells transduced with a control lentiviral vector in Once b-catenin is released from an inhibitory complex organoid cultures (Fig. 3a). Crypt cells were incubated with following Wnt signalling, it binds to DNA together with its Wnt3a to enhance transduction efficiency as previously partner, Tcf4, to induce Wnt pathway genes. In human reported2,3. Hmga1 expression was repressed in the organoids embryonic stem cells, HMGA1 induces expression of c-MYC transduced to express the shHmga1 (Fig. 3b). Strikingly, the crypt (ref. 16), a WNT/TCF4/b-catenin gene target, suggesting that cells with Hmga1 silencing formed very few organoids with Hmga1 could cooperate with Wnt/Tcf4/b-catenin to regulate the smaller projected surface areas and decreased bud number, while Wnt stem cell program. We therefore assessed expression of Wnt/ those transduced with control vector organized into typical 3D Tcf4/b-catenin target genes, including Axin2, Ascl2, b-catenin, structures and generated new buds (Fig. 3a–c; Supplementary Cd44, c-Myc, Ephb2, Ets, Tcf4 and Prom-1. Expression of all of Fig. 5a). To rule out any potential nonspecific toxicity from the these genes, excluding Prom-1, was increased by 42 to 6-fold in shHmga1 vector, we also tested an inducible, shRNA lentiviral Hmga1 ISCs versus WT control ISCs (Fig. 4e). Together, these vector targeting Hmga1 (inducible-shHmga1), which is linked to a results indicate that Hmga1 not only activates Wnt target genes gene encoding red fluorescent protein (RFP); both RFP and together with Tcf4/b-catenin, but also amplifies Wnt signals by shHmga1 are induced by doxycycline. Crypt cells transduced with inducing expression of Wnt agonist receptor genes. the lentivirus, but not induced to express the shHmga1 vector, To better define the role of Hmga1 in regulating Wnt organized into 3D organoids with buds similar to what we signalling, we cultured Hmga1 organoids in the absence of the observed in WT crypt cultures (Fig. 3d; top panels). As expected, Wnt receptor agonist, R-spondin 1 (R-spo1), which is essential the uninduced organoids (cultured without doxycycline) do not for organoid formation in this 3D culture system. R-spo1 is express RFP. In contrast, those organoids in which the shHmga1 secreted by intestinal stromal cells and binds to the Lgr5 receptor

NATURE COMMUNICATIONS | 8:15008 | DOI: 10.1038/ncomms15008 | www.nature.com/naturecommunications 5 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms15008

a b Day 1 Day 5 Day 10 ** 1.2 1 0.8

mRNA 0.6 Cont 0.4 0.2 Hmga1 0 c Cont KD 8 ** ) 2 m μ

KD 6 ( 4 4 **

2 *

organoid ×10 0

Projected surface area per Projected surface Cont KD Cont KD Cont KD d Day 1 Day 3 Day 5 Day 1 Day 5 Day 10

e Dox(–) Cont Dox(+) KD

1.2 * ** ** 1 Dox (–) Cont Dox 0.8

mRNA 0.6 0.4

RFP RFP RFP Hmga1 0.2 0 Day 1 Day 3 Day 5 f 10 )

2 **

m 8 Dox (+) KD Dox ** μ ( 4 6

RFP RFP RFP 2 organoid ×10 0 Projected surface area per Projected surface Day 1 Day 3 Day 5

Figure 3 | Hmga1 deficiency disrupts ISC function in 3D organoids. (a) Representative images of organoid formation after silencing Hmga1 by delivery of lentiviral vector expressing shRNA targeting Hmga1 (knockdown or KD) or control lentivirus (Cont) are shown. Scale bars, 50 mm. (b) Relative Hmga1 expression (mean±s.d.) was assessed from two experiments performed in triplicate (via qPCR) in organoids transduced with lentivirus (Cont versus KD) at day 10. Gapdh was used to control for loading. **Po0.01; two-tailed Student’s t-test. (c) PSA (mean±s.d.) was assessed in organoids transduced with lentivirus (Cont versus KD) at the indicated time points (n ¼ 20 organoids per group). *Po0.05, **Po0.01; two-tailed Student’s t-test. (d) Representative image of organoid formation±knockdown of Hmga1 via an inducible lentiviral vector expressing red fluorescent protein (RFP) and shRNA targeting Hmga1 are shown. Doxcyline (0.5 mgmlÀ 1) was used to induce RFP and shRNA. Scale bars, 50 mm. (e) Relative Hmga1 expression (mean±s.d.) in organoids with or without induction of lentivirus shRNA targeting Hmga1 was assessed from two experiments performed in triplicate (via qPCR); Gapdh was used to control for loading. *Po0.05, **Po0.01; two-tailed Student’s t-test. (f) PSA (mean±s.d.) of organoids with or without induction of Hmga1 knockdown at the indicated time points are shown (n ¼ 20 organoids per group from two different virus transduction experiments). **Po0.01; Mann–Whitney test. to activate Wnt signalling39,40. We discovered a striking were transduced with the Wnt reporter lentiviral vector, which difference in the response of Hmga1 organoids to the absence includes a puromycin resistance gene. After selection for stable of R-spo1 as compared to the WT controls: Hmga1 organoids integration of the Wnt reporter construct by puromycin, the continued to survive and proliferate for over 2 weeks, while the cultured cells were transduced with either control retrovirus or WT organoids lost their 3D structural organization and retrovirus expressing human HMGA1. In both cell types, ultimately died by day 5 (Supplementary Fig. 7). By 3 weeks, HMGA1 activated Wnt reporter activity, while retroviral however, the Hmga1 organoids also stopped proliferating and lost delivery of the control virus failed to do so (Fig. 4f). These their 3D organization, suggesting that Hmga1 overexpression experiments further support the role of HMGA1 in activating partially rescues loss of Wnt signalling via R-spo1, but does not Wnt signalling. completely recapitulate or bypass Wnt. To further test the link between Hmga1 and Wnt, we treated To further investigate the link between HMGA1 and Wnt, we crypt cells with the Wnt inhibitor, C59, which blocks Wnt- determined whether cells overexpressing HMGA1 have enhanced mediated transcription and cell proliferation by inhibiting Wnt reporter activity using an established Wnt reporter construct porcupine (PORCN), a protein required for Wnt palmitoylation, that includes seven canonical Tcf4/b-catenin-binding sites in secretion and biological activity. Crypt cells isolated from Hmga1 tandem upstream of a luciferase reporter gene46. Both HEK 293 or WT small intestine were selected for small crypt size and cell human embryonal kidney cells and Caco-2 colon cancer cells number by passage through a 40 mM filter. The Hmga1 crypts

6 NATURE COMMUNICATIONS | 8:15008 | DOI: 10.1038/ncomms15008 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/ncomms15008 ARTICLE

abWTHmga1 WT Hmga1 -catenin β

c FUGW FUGW-Hmga1 -catenin β

e d 10 ** 6 8 ** ** 6 ** 4 ** ** ** ** ** 4 ** ** ** 2 ** 2 Relative expression

Relative expression 0 0 Fzd5 Fzd7 Lrp5 Lrp6 Lgr5 Tcf4 Ets2 Axin2 Cd44 Ascl2 c-Myc catenin Ephb2 β- Prom-1

f g 20 HEK 293 20 Caco2 Day 3 Day 8 Day 14 **

10 10 ** WT

** ** Luciferase activity 0 0 Hmga1 PMX PMX PMXA1 Wnt3A PMXA1 Wnt3A h 0.03 WT 0.02 * *

0.01 IWP-2 C59 efficiency (%) Colony forming 0.00 WT Hmga1 WT Hmga1 Hmga1 C59 IWP-2

Figure 4 | Hmga1 amplifies Wnt signalling in ISCs. (a) Representative organoids after exposure to Wnt3a (10 ng ml À 1) are shown. Hmga1 organoids form large, cyst-like spheres (standard (top) and fluorescent (bottom) microscopy). Scale bars, 50 mm. (b) b-Catenin IHC staining is shown in small intestines from WT and Hmga1 mice. Scale bars, 20 mm. (c) b-Catenin IHC staining is shown in organoids engineered to express control lentivirus (FUGW; left) or Hmga1 (FUGW-Hmga1; right). Scale bar, 20 mm. (d) Relative expression (mean±s.d.) of genes encoding Wnt agonist receptors was ascertained by qPCR in GFP þ ISCs from WT (blue) or Hmga1 transgenic mice (orange) in two experiments performed in triplicate. Gapdh was used to control for loading. **Po0.01; two-tailed Student’s t-test. (e) Relative expression (mean±s.d.) of genes encoding Wnt/Tcf4/b-catenin transcriptional targets was ascertained by qPCR in GFP þ ISCs from WT or Hmga1 transgenic mice in two experiments performed in triplicate. Gapdh was used to control for loading. **Po0.01; two-tailed Student’s t-test. (f) HEK 293 (left) and Caco2 (right) cells infected with the synthetic Wnt reporter construct containing seven optimal Tcf4/b-catenin-binding sites showed activation in cells overexpressing Hmga1. Purified Wnt3A (100 ng ml À 1) protein was used as a positive control. Bars show mean luciferase activity±s.d. from two experiments performed in triplicate. **Po0.01; two-tailed Student’s t-test. (g) Representative image of GFP þ ISCs from WT or Hmga1 mice are shown in 3D culture with Wnt porcupine inhibitors IWP-2 (0.5 mM) and C59 (0.5 mM). Scale bars, 50 mm. (h) Relative colony-forming efficiency (mean±s.d.) from three experiments is shown with GFP þ ISCs from WTor Hmga1 mice in 3D culture with the Wnt porcupine inhibitors (IWP-2, C59). *Po0.05; two-tailed Student’s t-test. generated organoids and buds in C59 (0.25 mM) with a greater with vehicle alone. In contrast, the Hmga1 crypt cells were projected surface area than WT crypt cells cultured with C59 relatively impervious to C59; by day 6, there was no difference in (Supplementary Fig. 8). In addition, the mean projected surface projected surface area of Hmga1 organoids compared to those area of WT organoids cultured in C59 was reduced by days 6 and cultured with vehicle alone, although the projected surface area 10 compared to WT organoids cultured under control conditions decreased by day 10 when cultured with C59 compared to those

NATURE COMMUNICATIONS | 8:15008 | DOI: 10.1038/ncomms15008 | www.nature.com/naturecommunications 7 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms15008 cultured with vehicle control. This blunted response of Hmga1 Paneth cells, Defcr-rs51, and found that it is increased in the organoids to Wnt inhibition could result from the ability of Hmga1 organoids compared to control organoids (Fig. 5j). Hmga1 to amplify Wnt signalling, thereby mitigating the effects Together, these results indicate that Hmga1 induces Paneth cell of Wnt inhibition. Alternatively, or in addition to amplifying niche formation and expansion, which could help to promote ISC Wnt signalling, the Hmga1 organoids could dampen the effects of maintenance and expansion. Wnt inhibition by providing a buffer as a result of an increased size Because terminal differentiation to a Paneth cell requires Sox9 and cell number. Finally, an increase in the number of Paneth cells, (refs 47–49), we hypothesized that Hmga1 fosters Paneth cell which secrete Wnt agonists, could mitigate the effects of Wnt expansion by upregulating Sox9 expression. HMGA1 induces the inhibition. To distinguish between these possibilities, we isolated SOX family member, SOX2, in human embryonic stem cells16; single, purified Lgr5 þ ISCs from transgenic or WT mice and further, both human and mouse SOX9/Sox9 have similar AT-rich cultured them in the presence of two different Wnt PORCN regions and predicted Hmga1 DNA-binding sites in the 50 inhibitors, C59 and IWP-2. Strikingly, the Hmga1 ISCs proliferated untranslated region. Sox9 is also a b-catenin/Tcf4 target gene. We and formed 3D, cyst-shaped organoids for up to 10 days, albeit at a found that Hmga1 upregulates Sox9 expression in the Lgr5 þ slower rate between days 6–10 compared to purified Hmga1 ISCs ISCs as compared to WT Lgr5 þ ISCs (Fig. 6a). To determine cultured without inhibitor (Fig. 4g,h). In contrast, the WT ISCs whether Hmga1 directly induces Sox9 gene expression, we generated very few organoids when cultured with C59. Moreover, performed chromatin immunoprecipitation (ChIP) in intestinal no organoids formed from WT ISCs cultured in IWP-2 (Fig. 4g,h). crypt cells from WT mice. We discovered that Hmga1 binds Together, these findings demonstrate that Hmga1 ISCs have a directly to the Sox9 promoter at two conserved sites (Fig. 6b; blunted response to Wnt inhibitors, which is consistent with their Supplementary Fig. 9a,b). Enrichment for Hmga1 binding was ability to amplify Wnt signals. greatest at the proximal site (denoted site 1) and significant, albeit lower, at the second most proximal site (site 2; Po0.01 by two- tailed Student’s t-test). Hmga1 binding was not enriched at the Hmga1 expands the Paneth cell niche and upregulates Sox9. distal site (site 3). For a positive control, we used a histone H3 Paneth cells are terminally differentiated epithelial cells derived antibody, which showed significant enrichment (Po0.01 by two- from ISCs and located at the base of intestinal crypts47–50. They tailed Student’s t-test) at all promoters tested. For a negative support ISC survival by secreting Wnt3a and other factors, thus control, we used the IgG antibody, which had minimal binding to providing an epithelial niche for ISCs. To determine whether chromatin at all promoter regions tested. As a second negative Hmga1 alters the Paneth cell niche in our transgenic models, control, we used the murine Hprt gene; this promoter region was we stained for lysozyme using alkaline phosphatase, which previously shown to be negative for Hmga1 occupancy. We marks Paneth cells. There was a marked increase in lysozyme confirmed that Hmga1 does not occupy the Hprt promoter stain (Fig. 5a), which was confirmed quantitatively by (Fig. 6b). Next, we performed transfection experiments to immunohistochemistry (IHC) pixels (Fig. 5b). To determine determine whether Hmga1 transactivates the Sox9 promoter whether the increase in lysozyme stain corresponds to an increase reporter construct containing the Hmga1-binding site 1, which in Paneth cell number, Paneth cells were enumerated in mouse had the greatest enrichment for Hmga1 binding (Supplementary intestinal tissue after controlling for the region of the intestine Fig. 9a–c). As a control, we included a Sox9 promoter reporter (Fig. 5c; n ¼ 100 crypts per group from duodenal epithelium). construct lacking this Hmga1-binding site (Supplementary There was a marked increase in Paneth cell number per crypt, Fig. 9b,c). We found that Hmga1 induced the Sox9 promoter which paralleled the increase in lysozyme stain by IHC. Next, we construct that included site 1, while there was no induction of the co-stained the duodenal epithelium with EpCAM to demarcate promoter construct lacking this site (Supplementary Fig. 9d). As cell borders, DAPI to indicate individual nuclei, and lysozyme an additional control, we included a dominant-negative HMGA1 such that the frequency of cells staining positive for lysozyme construct with mutations in the second AT-hook DNA-binding could be enumerated within the crypts (Fig. 5d). There was also a domain, which no longer binds to DNA16. The dominant- marked increase in Paneth cell frequency (Fig. 5e; n ¼ 50 crypts negative mutant also failed to transactivate either Sox9 promoter per group from duodenal epithelium), consistent with an construct (Supplementary Fig. 9d). To further validate this expansion in the Paneth cell niche in the Hmga1 transgenic relationship between Hmga1 and Sox9, we assessed Sox9 mRNA duodenal epithelium as compared to WT control duodenal in organoid cells transduced to overexpress Hmga1 and found epithelium. To further confirm the increase in Paneth cells by that Sox9 was induced by Hmga1 (Supplementary Fig. 9e). In Hmga1, we assessed Paneth cell number in Hmga1 organoids contrast, we also found that Sox9 was repressed in organoids with using two different approaches. First, Paneth cells have a Hmga1 knockdown following transduction with shRNA targeting distinctive granular appearance on phase contrast microscopy. Hmga1 compared to controls (Supplementary Fig. 9f). We therefore compared the mean number of granular cells per To determine whether cells with high levels of Hmga1 protein bud in WT and Hmga1 organoids after controlling for bud size by also have high levels of Sox9 protein, we assessed Sox9 protein dividing the Paneth cell number by the projected surface area per and found that it is increased in the crypts of the Hmga1 bud (Fig. 5f,g; n ¼ 35 buds per group). Similar to the Hmga1 transgenic mice compared to WT controls by IHC (Fig. 6c). To mouse intestinal tissue, there was a significant (Po0.00001; quantitatively assess Hmga1 and Sox9 proteins, we performed Mann–Whitney test) increase in Paneth cell number per bud western blot analysis of crypt cells from WT and transgenic mice; projected surface area in the Hmga1 organoids (Fig. 5g). Second, the Hmga1 transgenic crypt cells had an increase in both Hmga1 we assessed Paneth cell number by staining with lysozyme, which and Sox9 protein compared to WT crypt cells (Fig. 6d,e). Next, appeared to be increased in organoids transduced to express we compared adjacent sections stained for Hmga1 or Sox9 and Hmga1 compared to control organoids (Fig. 5h). To found that the majority of cells with detectable Hmga1 at the base quantitatively assess the frequency of Paneth cells per total bud of the crypts also have detectable Sox9 protein in both WT and cell number, we co-stained with EpCAM, DAPI and lysozyme. transgenic mice (Fig. 6f,g). In WT mice, 470% of cells that There was an increase in Paneth cell frequency using this stained positive for Hmga1 also stained positive for Sox9. In the approach, similar to our results with intestinal tissue and granular transgenic mice, which have more cells with Hmga1 staining and cell number (Fig. 5h,i). Finally, to further corroborate this result, higher levels of Hmga1, we found that 495% of cells that stained we assessed expression of a transcript that is highly specific for positive for Hmga1 also stained positive for Sox9 (Fig. 6f,g).

8 NATURE COMMUNICATIONS | 8:15008 | DOI: 10.1038/ncomms15008 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/ncomms15008 ARTICLE

bc a ** 10 WT Hmga1 ** 70,000 60,000 8 50,000 6 40,000 30,000 4 per crypt 20,000 2 Paneth cell number Paneth Lysozyme IHC pixels Lysozyme 10,000 0 0 WT Hmga1 WT Hmga1

defWT Hmga1 ** WT Hmga1 60

20 cells per crypt % Lysozyme positive % Lysozyme 0

WT Hmga1

ghFUGW FUGW-Hmga1 ij

5 ** 20 2.5 4 ** –4 2 ** 3 15 1.5 2 10 mRNA

bud area ×10 bud 1 1 cells per bud

5 Defcr-rs Granular cell number per cell number Granular 0.5 0 positive % Lysozyme 0 WT 0 Hmga1 FUGW Hmga1 FUGW Hmga1

Figure 5 | Hmga1 expands the Paneth cell niche. (a) IHC for lysozyme is shown from small intestines of WTand Hmga1 transgenic mice. Scale bar, 20 mm. (b) Lysozyme IHC pixels (mean±s.d.; imagine pro-plus 6.0 software) is shown in small intestinal cross-sections (n ¼ 78 crypts per group; 3 mice per genotype). **Po0.01; Mann–Whitney test. (c) Paneth cell number (mean±s.d.) per crypt (n ¼ 100 crypts per group; 3 mice per genotype). **Po0.00001; Mann-Whitney test. (d) Immunofluorescent co-staining for lysozyme, Ep-CAM and DAPI is shown from small intestines of WTand Hmga1 transgenic mice. Scale bars, 20 mm. (e) Paneth cell frequency (mean±s.d.) was obtained by dividing the Paneth cell number by total cell number per crypt in WTand Hmga1 intestine (n ¼ 50 crypts per group; 3 mice per genotype). **Po0.00001; Mann–Whitney test. (f) Granular cells in WTand Hmga1 organoid buds are shown via phase contrast microscopy. Scale bars, 50 mm. (g) Granular cells per bud PSA (mean±s.d.) from WTand Hmga1 organoids are shown (n ¼ 35 buds per group). **Po0.00001; Mann–Whitney test. (h) IHC for lysozyme (top) and immunofluorescent co-staining for lysozyme, Ep-CAM and DAPI (bottom) are shown in organoids from WT or Hmga1 mice. Scale bars, 50 mm. (i) Paneth cell frequency per bud (mean±s.d.) was calculated by dividing the Paneth cell number by the total bud cell number (n ¼ 30 organoids per group). **Po0.00001; Mann–Whitney test. (j) Relative expression (mean±s.d.) of Defcr-rs, the Paneth cell-specific transcript, is shown from two experiments performed in triplicate from organoids from WTand Hmga1 mice. Gapdh was used to control for loading. **Po0.01; two-tailed Student’s t-test.

While Sox9 is required for Paneth cell differentiation47–51,itis proliferation in vivo in intestinal epithelium49, we observed that not known whether increased levels of Sox9 lead to expansion in the organoid growth (estimated by projected surface area) was Paneth cells. To test this, we transduced WT organoids with a decreased in organoids with Sox9 induction compared to lentivirus expressing Sox9 following induction by doxycycline uninduced organoids (Fig. 7d,e). We also found that the (Fig. 7). Immunoﬂuorescent staining showed increased Sox9 projected bud surface area was decreased with Sox9 induction protein in the doxycycline-induced organoids compared to (Fig. 7f). Next, we compared Paneth cell frequency in Sox9- uninduced control organoids cultured for the same period of induced and uninduced organoids using two different time (Fig. 7a). Sox9 protein (by western blot analysis) and mRNA approaches. First, we used phase contrast microscopy to (by qPCR) were both increased relative to uninduced organoids identify Paneth cells morphologically by the distinctive (Fig. 7b,c). Consistent with the observation that Sox9 decreases granules, which are characteristic of these cells. The relative

NATURE COMMUNICATIONS | 8:15008 | DOI: 10.1038/ncomms15008 | www.nature.com/naturecommunications 9 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms15008

c WT Hmga1 ab Anti-IgG ** 25 3 Anti-Hmga1 ** Anti-H3 20 2 **

mRNA ** 15 ** 1 % Input **

Sox9 ** 10 5 0

WT 0 Hmga1 S1 S2 S3 Hprt

d e Sox9 Hmga1 2 2.5 * * kDa WT Hmga1 70 2 1.5 Sox9 55 1.5 25 1 Hmga1 1 15 0.5 0.5 35 Gapdh density Relative 0 0 WT Hmga1 WT Hmga1

fgAnti-Hmga1 Anti-Sox9 ** ** ** 35 30 25 WT H 20 S 15 10 per crypt 5

Positive cell number Positive 0

Hmga1 S WT WT WT H Hmga1 Hmga1 Hmga1 Hmga1(+) Sox9(+) Double(+)

Figure 6 | Hmga1 directly induces Sox9 expression. (a) Relative Sox9 mRNA (mean±s.d.) by qPCR in GFP þ ISCs from WT and Hmga1 mice is shown from two experiments performed in triplicate. Gapdh was used to control for loading. **Po0.01; Student’s t-test. (b) Hmga1 binding to the Sox9 promoter by ChIP is shown from mouse crypt cells. IgG antibody and the Hprt promoter region were used as negative controls; histone H3 was a positive control. Bars show mean enrichment±s.d. from two experiments performed in triplicate; **Po0.01; two-tailed Student’s t-test. (c) Sox9 IHC staining (brown) is shown in intestinal cross-sections from WT and Hmga1 transgenic mice. Scale bars, 20 mm. (d) Western blots for Sox9, Hmga1, and Gapdh (loading control) were performed from freshly isolated crypt cells from WTand Hmga1 transgenic mice. Western blots were done three times; a representative blot is shown. Size markers (kDa) are indicated. (e) Densitometry analysis±s.d. was performed on repeat western blots. *Po0.05; two-tailed Student’s t-test. (f) Sox9 and Hmga1 IHC staining (brown) is shown in adjacent sections of small intestine from WTand Hmga1 mice. Arrows show representative staining for Hmga1 (H) or Sox9 (S); triangles show Hmga1 transgenic lymphocytes which have high levels of Hmga1 (but not Sox9). Scale bars, 50 mm. (g) Mean number of cells±s.d. staining positive for Hmga1, Sox9 or both were identiﬁed by IHC in adjacent cross-section on slides of small intestinal crypts from WT and Hmga1 mice (nZ21 crypts per group.) **Po0.01 by Mann–Whitney test.

Paneth cell number per bud-projected surface area was increased be other factors that contribute to Paneth cell differentiation. in the Sox9-induced organoids compared to uninduced control Together, our results demonstrate that Hmga1 drives self-renewal organoids (Fig. 7g,h). Second, we used lysozyme stain to identify and expansion of ISCs in addition to helping to establish an Paneth cells, together with EpCAM stain (to delineate cell epithelial stem cell niche through Paneth cell differentiation. membranes)50 and DAPI (to demarcate nuclei). There was also an increase in the frequency of Paneth cells per total bud cell HMGA1 and SOX9 in colonic epithelium and colorectal cancer. number in Sox9-induced organoids (n450 buds per condition To determine whether our ﬁndings in mouse intestine are rele- from 450 organoids per condition; Fig. 7i,j). To further vant to humans, we assessed expression of HMGA1 and SOX9 in substantiate these results, we assessed expression of a transcript human intestinal epithelium. Using the Cancer Genome Atlas that is highly speciﬁc for Paneth cells, Defcr-rs51, and showed that (TCGA), we found that HMGA1 and SOX9 are positively corre- it is also increased in the organoids overexpressing Sox9 (Fig. 7k). lated in normal colonic epithelium (P ¼ 0.008, r ¼ 0.52; Fig. 8a). These results suggest that Hmga1-mediated induction of Sox9 We also stained human small intestine, and found that HMGA1 helps to expand the Paneth cell niche, although there are likely to localizes to the columnar basal cells within the crypts where ISCs

10 NATURE COMMUNICATIONS | 8:15008 | DOI: 10.1038/ncomms15008 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/ncomms15008 ARTICLE

abSox9-Fuw Sox9-Fuw c Dox(–) Dox(+) 4 ** 4 ** kDa Dox(–)Dox(+) 70 3 3 55 Sox9 2 2 mRNA

Gapdh 1 1 35 Sox9 Relative density Relative 0 0 Dox(–) Dox(+) Dox(–) Dox(+) EpCAMSox9DAPI EpCAMSox9DAPI

d Sox9-Fuw Sox9-Fuw e f Dox(–) Dox(+) 20 4 ) ** ** 2

m 15 )

μ 3 2 ( 4 m μ 10 ( 2 4 5

×10 1 area ×10 Projected organoid 0 area Projected bud 0 g Dox(–) Dox(+) Dox(–) Dox(+) h

5 4 ** 4

1 projected bud area ×10 projected bud

Granular cell number per cell number Granular 0 Dox(+) Dox(–)

ijk EpCAM EpCAM 25 Lysozyme Lysozyme **6 ** DAPI DAPI 20 5

15 4

mRNA 3 10 2 cells per bud Defcr-rs 5 1 % of lysozyme positive 0 0 Dox(–) Dox(+) Dox(–) Dox(+)

Figure 7 | Paneth cell number increases in organoids overexpressing Sox9. (a) Immunofluorescent stain for Sox9 (red), DAPI (blue) and EpCAM (green) is shown in organoids transduced with lentivirus expressing doxycycline-dependent inducible Sox9 (Sox9-Fuw). Sox9 (Sox9-Fuw Dox( þ )) was induced by doxycycline (1 mgmlÀ 1) for 5 days; controls were cultured for 5 days without doxycycline. (b) Representative western blot for Sox9 in control and induced organoids is shown. Densitometry±s.d. was performed on three repeat western blots. **Po0.01; two-tailed t-test. (c) Sox9 mRNA (mean±s.d.) was assessed by qPCR from two experiments performed in triplicate in control and induced organoids. Gapdh was used to control for loading. **Po0.01; two- tailed Student’s t-test. (d) Typical organoids from uninduced controls (left) and induced (right) cultures are shown. (e) PSA±s.d. of organoids (n ¼ 52 per group) are shown. **Po0.01; two-tailed Student’s t-test. (f) PSA±s.d. of buds (n ¼ 52 per group) are shown. **Po0.01; two-tailed Student’s t-test. (g) Paneth cells identified as granular cells using phase contrast microscopy within individual buds are shown. (h) Granular cell frequency (mean±s.d.) was estimated by dividing the total granular cell number by the total bud PSA (n ¼ 52 per group). **Po0.01; Mann–Whitney test. (i) Paneth cells were identified based on lysozyme stain (red) in cells delineated by EpCAM (green) for cell membrane and (DAPI) for nuclei. (j) Paneth cell frequency (mean±s.d.) was ascertained by dividing the number of lysozyme staining cells (red) delineated by EpCAM (green) by the total cell number (DAPI) per bud (n ¼ 35 per group). **Po0.01; Mann–Whitney test. (k) Relative expression of Defc-rs (mean±s.d.) was compared in uninduced control and Sox9-induced organoids by qPCR from two experiments performed in triplicate. Gapdh was used to control for loading. **Po0.01; two-tailed Student’s t-test. Scale bars, 50 mm.

NATURE COMMUNICATIONS | 8:15008 | DOI: 10.1038/ncomms15008 | www.nature.com/naturecommunications 11 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms15008

a 16 ** P<0.008 ** P<0.000001 14 ** P<0.000001 15 11 13 14 12 10 13 11 expression 9 12 expression 10 in non-malignant 11 9 SOX9 intestinal epithelium 8 HMGA1

SOX9 SOX9 10 8 10 10.5 11 11.5 12 Control Cancer Control Cancer HMGA1

Key: Lgr5+ISC Transit amplifying cell Goblet cell Differentiated enterocyte Paneth-like cell Aberrant Aberrant Paneth cell reprogramming reprogramming Neoplastic/malignant cells to a neoplastic, to a neoplastic, stem-like state stem-like state

Sox9 Sox9 Sox9 Sox9

Hmga1 Hmga1 Inflammation Normal Normal Inflammation injury homeostasis homeostasis injury Wnt genomic lesions Wnt genomic lesion

b Colon c Small intestine

Figure 8 | Hmga1 in normal intestinal homeostasis and reprogramming to neoplasia. (a) There is a significant positive correlation (r ¼ 0.51; P ¼ 0.008) between HMGA1 and SOX9 in control, non-malignant large intestinal epithelium by Spearman’s correlation (n ¼ 26). There is also a highly significant upregulation of both HMGA1 and SOX9 in colorectal cancer (n ¼ 293, Po0.000001). Boxplots, scatterplots and Spearman rank-based correlations were calculated using R statistical software. (b) Model depicting normal ISC function and tissue homeostasis in the large intestine under conditions of tightly regulated Hmga1 expression and Wnt signalling (left). Intestinal epithelium homeostasis is disrupted (right) in the setting of aberrant Hmga1 expression, leading to expansion in the ISC compartment, excessive Wnt signalling, and abnormal proliferation, culminating in epithelial reprogramming to neoplastic, transformed cells. (c) Model depicting normal ISC function and tissue homeostasis in the small intestine under conditions of tightly regulated Hmga1 expression, Paneth cell differentiation and Wnt signalling (left). Expansion in the ISCs and Paneth cell niche occurs when homeostasis is disrupted (right) in the setting of aberrant Hmga1 expression. This contributes to amplified Wnt signalling, abnormal proliferation, epithelial reprogramming and neoplastic transformation. are located (Supplementary Fig. 10). Because HMGA1 is over- in these settings had been poorly understood. Our studies reveal a expressed in diverse epithelial cancers and correlates with cancer role for Hmga1 in both stem cell self-renewal and establishment stem cell properties in experimental models, we also sought to of a stem cell niche within small intestinal crypts. The HMGA determine whether HMGA1 and SOX9 are co-regulated in col- gene family includes HMGA1 (on chromosome 6p21) and orectal cancer. Strikingly, both HMGA1 and SOX9 are markedly HMGA2 (on chromosome 12q15), both of which are highly upregulated in human colorectal cancer (Po0.0000001), expressed during embryonic development, but with low or although their expression was not correlated in this setting undetectable levels in differentiated tissues13,16,25,26. Aberrant (Fig. 8a). Interestingly, the relative expression and amplitude of expression of HMGA1 occurs in most poorly differentiated HMGA1 mRNA levels are greater than that of SOX9 (Fig. 8a). human cancers, including gastrointestinal cancers such as colon, These data support a model whereby HMGA1 and SOX9 are gastric, pancreatic and esophageal cancers, and high levels crucial for normal ISC function, and both become upregulated in correlate with poor outcomes in diverse tumours13,30–35. carcinogenesis (Fig. 8b,c). These findings also suggest that upre- Notably, a prior study in colorectal cancer identified HMGA1 gulation of both HMGA1 and SOX9 beyond a threshold may be among the genes most enriched in cancer relative to adjacent, necessary for early reprogramming of an epithelial cell to a non-malignant tissue33. HMGA1 is also required for properties neoplastic cell, while further increases in HMGA1 could drive attributed to cancer stem cells, including tumour initiator cells, tumour progression. growth as 3D spheres and metastatic progression14,18. In contrast, HMGA2 overexpression occurs primarily in benign tumours of Discussion mesenchymal origin31, as well as a subset of malignant HMGA1 expression has been identified among genes most tumours52–55. The dual role for Hmga1 in normal development enriched in embryonic and adult stem cells, although its function and poorly differentiated cancers suggests that it regulates cell fate

12 NATURE COMMUNICATIONS | 8:15008 | DOI: 10.1038/ncomms15008 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/ncomms15008 ARTICLE decisions, although a detailed understanding of molecular to ductal cells during carcinogenesis63. Similarly, HMGA1 could mechanisms involved in these processes was previously unknown. induce SOX9 and enforce Wnt signalling to drive stem cell Here we show that Hmga1 amplifies Wnt signalling to drive properties and reprogram intestinal epithelial cells during self-renewal and ISC expansion. Hmga1 not only upregulates carcinogenesis. Although HMGA1 and SOX9 are positively Wnt agonist receptor genes, but also enhances expression of correlated in normal colonic epithelium and both become genes downstream of WntTcf4/b-catenin. Our results, together upregulated in cancer, it is not surprising that the correlation is with the prior finding that Tcf4 binds to the HMGA1 promoter in lost in colorectal cancer. This is consistent with the large body of colorectal cancer cells56, suggest that Hmga1 is involved in a evidence demonstrating that colon cancer is initiated by the ‘feed-forward’ loop, whereby Tcf4/b-catenin induces Hmga1, progressive acquisition of mutations in the APC pathway, which leading to enhanced Wnt signalling. The expansion of ISCs in our include genes upstream of SOX9 and the SOX9 gene itself10,11. transgenic mouse model was recapitulated in vitro in organoid Thus, there are multiple distinct mechanisms that could give rise cultures, which depend on Hmga1 for organization into 3D to overexpression of HMGA1 and SOX9 in colorectal cancer, and structures and bud formation. Our transgenic mouse and SOX9 may depend, at least in part, on genetic lesions distinct organoid models provide valuable tools to further dissect from HMGA1 overexpression in this setting. Together, our work downstream pathways regulated by Hmga1 in ISCs. In a not only reveals a novel function for Hmga1 in intestinal murine model of gastric cancer, Wnt signalling also upregulates homeostasis through self-renewal of ISCs and Paneth cell Hmga1 (ref. 57). We are the first to show that Hmga1 enhances differentiation, but also sheds light on possible mechanisms Wnt/b-catenin signalling at multiple levels in the pathway, involved in Hmga1-mediated neoplastic transformation and consistent with a feed-forward amplification loop, whereby Wnt intestinal carcinogenesis. induces Hmga1, which in turn, upregulates Wnt/b-catenin signalling (Fig. 7). Methods Our work also uncovered a role for Hmga1 in Paneth cell Mouse models. The Hmga1 transgenic construct and mice have been previously differentiation, mediated in part, through Sox9. Paneth cells described35,36. Female Lgr5-eGFP-IRES-CreERT2 mice6 (Jackson Labs) were constitute an epithelial niche, providing Wnt3a and other signals crossed with male Hmga1a transgenics35. All mice were on the C57Bl6 background. Animal experiments were conducted in accordance with our to maintain ISCs and permit self-renewal within the crypts. Cues institutional Animal Care and Use Committee (protocol# MO14M187). All mice from the epithelial and stromal niche are likely to help govern were housed in a sterile environment where they had free access to food and water whether Hmga1 induces Sox9 to drive Paneth cell differentiation as outlined in our institutional guidelines. The ages of the mice were 2–4 months; or other Wnt signals to drive self-renewal. Because adult stem B50% were male. Experiments comparing wild-type to transgenic mice were cells can divide asymmetrically, Hmga1 could promote ISC matched for age and gender in all cases. division to generate both an identical daughter cell and a differentiated Paneth cell, or even a more fully differentiated Crypt isolation. Crypts were isolated using an established protocol4–7,41. First, transit-amplifying cell. Recent studies indicate that R-spondin 1, intestines were flushed with phosphate-buffered saline (PBS) and incised longitudinally after which villi were removed mechanically by scraping. Sections another Wnt receptor agonist secreted by intestinal stromal cells, (1 cm) were incubated in EDTA (5 mM)/PBS for 15 min at 4 °C per fraction of 39,40 is required for ISC maintenance in mice . It remains to be epithelium. After incubation, the epithelium was separated by vigorous shaking seen whether Hmga1 also regulates R-spondin 1 in stromal niche and the remaining intestinal tissue was placed in a new tube for collection of cells. Of note, mesenchymal stem cells within the bone marrow subsequent fractions. After isolation, crypt cells were pelleted, passed through a 70 mm cell strainer (unless indicated otherwise), evaluated for purity niche also express high levels of HMGA1, although the functional microscopically and counted. For purification of single Lgr5-GFP þ ISCs, isolated 58 significance of HMGA1 in this setting is not yet known . Prior crypts were incubated in culture medium for 45 min at 37 °C, followed by studies indicate that quiescent, label-retaining cells that are trituration with a glass pipette. Dissociated cells were passed through a cell strainer responsible for intestinal epithelial regeneration following injury with a pore size of 20 mm. GFP þ and GFP À cells were sorted by FACS Aria also differentiate into secretory progenitors and Paneth cells59. model (BD Biosciences). Hmga1 may also function in this setting, although lineage-tracing experiments will be needed to test this. Organoid culture and re-plating assay. Mouse organoids were established and maintained from isolated crypts of the proximal small intestine or from Lgr5 þ In addition to establishing a role for Hmga1 in maintaining ISCs isolated as single cells4–8,41. The basic culture medium (advanced Dulbecco’s ISC and niche compartments, we also found a positive correlation modified Eagle’s medium/F12 supplemented with penicillin/streptomycin, between HMGA1 and SOX9 in normal human large intestinal 10 mmol l À 1 HEPES, 13 Glutamax, 13 B27 (all from Life Technologies), and epithelium. Moreover, both genes are upregulated in colorectal 1 mmol l À 1 N-acetylcysteine (Sigma)) was supplemented with 50 ng ml À 1 murine recombinant EGF (Peprotech), R-spondin 1 (1 mgmlÀ 1) and Noggin cancer, indicating that this pathway may contribute to human À 1 14,33 (10 ng ml ). Wnt inhibitors C59 and IWP-2 are commercially available (Abcam). intestinal carcinogenesis . A recent study in the Adenomatous Conditioned media was produced using HEK 293T cells stably transfected with þ / À Polyposis Coli (Apc) murine model of intestinal carcino- HA-mouse Rspo1-Fc (a gift from Calvin Kuo, Stanford University). Advanced genesis showed that Hmga1 is downstream of the miR-26 tumour DMEM/F12 media supplemented with penicillin/streptomycin, 10 mmol l À 1 suppressor60, suggesting that APC mutations in colorectal cancer HEPES, and 13 Glutamax was conditioned for 1 week. For the re-plating assay, an equal number of cells from each model were isolated by mechanical dispersion with lead to HMGA1 induction through loss of miR-26. HMGA1 is a 10 ml pipet after 3–5 days in culture. Cells were then counted and re-plated using 13 also upregulated in the setting of inflammation , and intestinal the conditions described above. carcinogenesis is frequently preceded by chronic inflammation 61 and injury . Thus, both inflammatory lesions and genetic Lentivirus and transduction. The FUGW control vector (Addgene plasmid alterations (APC inactivating mutations) could cooperate to # 14883) and FUGW-Hmga1 lentiviral vectors36,64 have been described14,16,18. induce HMGA1 during carcinogenesis. Paneth cell expansion also The shRNA interference plasmid for Hmga1 (#TRCN0000182169; the RNAi 14,16,18 occurs in mice with deletion in the Apc tumour suppressor62, and Consortium/TRC) has been described . The empty shRNA vector was used as a control. For inducible silencing, pTRIPz-Hmga1-shRNA linked to RFP reporter Hmga1 could contribute to this phenotype. Although there are no was engineered to be inducible by tetracycline or analogues (Tet-On) and produces discrete populations of Paneth cells outside of the proximal colon tightly regulated induction of shRNA expression in the presence of doxycycline in humans, lysozyme-expressing Paneth-like cells are found in (0.5 mgmlÀ 1). The annealed Hmga1-shRNA oligonucleotides were cloned into human adenomas62. Based on our findings, it is plausible that linearized pTRIPZ empty vector (Open Biosystems catalogue number RHS4750). For pTRIPz-Hmga1-shRNA transductions, puromycin (2 mgmlÀ 1) was added to HMGA1 induces formation of Paneth-like niche cells during media after 2–3 days for selection. For all lentiviral transductions, we modified an colon carcinogenesis in the intestinal epithelium. Studies in established protocol65 using magnetic nanoparticles (ViroMag R/L, OZ Bioscience, murine pancreatic cancer show that SOX9 reprograms acinar cells Inc.) and a magnetic plate (ViroMag R/L, OZ Bioscience, Inc., catalogue number:

NATURE COMMUNICATIONS | 8:15008 | DOI: 10.1038/ncomms15008 | www.nature.com/naturecommunications 13 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms15008

MF10000) to transduce crypt cells and organoids. Organoid fragments were seeded of single organoids were contrasted to exclude background fluorescence, maxima- with 150 ml transduction medium into 48-well plates. Virus was added with projected to create a single plane image per channel, after which a threshold was viroMag R/L solution for 15 min at room temperature (2,500–3,000 virus particles applied for specific signal within organoid boundaries. Projected monochromatic per cells) to cells. The cell culture plate was placed on the magnetic plate for surface area was then measured within each imaged organoid, and expansion rate 30–60 min in a 37 °C tissue culture incubator. Cells were then incubated overnight of Lgr5 þ cell population in WT and Hmga1 mice, separately, was calculated by at 37 °C. Organoid fragments and transduction media were then transferred to a normalization of day 6 projected area over day 0 projected area in the green 1.5 ml tube for centrifugation at 900 g for 5 min. The supernatant was discarded channel. and tubes containing the pellet was placed on ice for 5 min. Next, 120 ml of matrigel was added and the pellet was resuspended by pipetting slowly up and down. Drops (30 ml) of basement matrix–cell mixtures were seeded into a new 48-well plate and Cell culture. Caco-2 and HEK 293 cell lines (from ATCC) were cultured in high- incubated at 37 °C for 5–15 min until the basement matrix solidified. Media glucose DMEM, supplemented with 4 mM glutamine, 50 U ml À 1 penicillin, (ENRWntNic)6 was then added to each well and placed into a tissue culture 50 mgmlÀ 1 streptomycin (all from Invitrogen) and 10% fetal bovine serum. All cell incubator. Common ENR media6 was used and changed every 2–3 days beginning lines were routinely tested for mycoplasma contamination with the MycoAlert 4–6 days after the transduction. Mycoplasma Detection Kit (Lonza). For cell line authentication, short-tandem repeat analysis was performed on isolated genomic DNA with the GenePrint 10 System from Promega, and peaks were analysed using GeneMarker HID from Gene expression analysis and chromatin immunoprecipitation. Wnt signalling Softgenetics. Allele calls were searched against short-tandem repeat databases 14–16 molecule gene expression was detected by qPCR (see Supplementary Table 1 maintained by ATCC (www.atcc.org) and DSMZ (www.dsmz.de). for primer sequence). Reactions were performed using the SYBR Green PCR Master Mix (Applied Biosystems) with the ABI 7500 qRT–PCR machine. Gapdh mRNA was assessed as the loading control14. For ChIP, primers were designed Data availability. The TCGA RNA-seq data (Fig. 8a) have been deposited in EZID using the sequence data from MatInspector in silico transcription factor-binding (at the California Digital Library) with identifier: Broad GDAC Firehose site prediction algorithm66 with amplicon sizes of around 200 bp. Both positive stddata__2013_08_09 run, doi:10.7908/C17W6BB3. Relevant primary data are control antibodies (histone H3) and negative control antibodies (IgG) as well as a presented in the manuscript and supplementary files; all primary data are available negative control promoter sequence (from the Hprt gene) were included36,67. from the authors on request. (See Supplementary Table 2 for ChIP primer sequences and antibodies.) All experiments were done in triplicate and performed at least twice. References 1. Clevers, H. The intestinal crypt, a prototype stem cell compartment. Cell 154, Western blots . Total protein from organoid cultures or tissues was isolated using 274–284 (2013). RIPA buffer (Sigma, USA). Lysates were centrifuged and separated on a gradient 2. Koo, B. K. et al. Controlled gene expression in primary Lgr5 organoid cultures. gel (8–12%) via SDS–polyacrylamide gel electrophoresis, blotted onto a membrane Nat. Methods 9, 81–83 (2011). (polyvinylidene fluoride; Bio Rad, USA), and analysed using specific antibodies 3. Koo, B. K., Sasselli, V. & Clevers, H. Retroviral gene expression control in (Supplementary Table 3). Bands were visualized using enhanced chemilumines- primary organoid cultures. Curr. Protoc. Stem Cell Biol. 27, 5A.6.1-5A.6.8 cence (ECL Kit, Amersham Biosciences, USA). Images of all uncropped western (2013). blots can be found in Supplementary Fig. 11. 4. Sato, T. & Clevers, H. Growing self-organizing mini-guts from a single intestinal stem cell: mechanism and applications. Science 340, 1190–1194 (2013). Statistical analysis. Data from all experiments were assessed for normal 5. Koo, B. K. & Clevers, H. Stem cells marked by the R-spondin receptor LGR5. distribution (Ryan–Joyner and D’Agostino–Pearson tests), and when normally Gastroenterology 147, 289–302 (2014). distributed, compared using the two-tailed, Student’s t-test. The Mann–Whitney 6. Barker, N. et al. Identification of stem cells in small intestine and colon by test was used for data that were not normally distributed. Po0.05 was considered marker gene Lgr5. Nature 449, 1003–1007 (2007). significant. Experiments that were technical failures, such as in vitro cultures in 7. Schuijers, J. et al. Ascl2 acts as an R-spondin/Wnt-responsive switch to control which controls did not grow, were excluded from the statistical analysis. Sample stemness in intestinal crypts. Cell Stem Cell 16, 158–170 (2015). sizes were chosen empirically based on previous experiments; no statistical 8. Cheung, E. C. et al. TIGAR is required for efficient intestinal regeneration and methods were used to predetermine sample sizes. Investigators were not blinded. tumorigenesis. Dev. Cell 25, 463–477 (2013). Mice were analysed based on genotype; randomization was not performed. 9. Farin, H. F. et al. Visualization of a short-range Wnt gradient in the intestinal stem-cell niche. Nature 530, 340–343 (2016). 10. Fearon, E. R. Molecular genetics of colorectal cancer. Annu. Rev. Pathol. 6, TCGA gene expression analysis. Transcript abundance for HMGA1 and SOX9 479–507 (2011). from human RNA-Seq data sets obtained from 293 primary colorectal cancers and 11. Vogelstein, B. et al. Cancer genome landscapes. Science 339, 1546–1558 (2013). 26 non-malignant colonic epithelial samples were assessed using RNA-Seq expectation maximization (RSEM) normalization. IluminaGa RNA-Seq data are 12. Reeves, R., Edberg, D. D. & Li, Y. Architectural transcription factor HMGI(Y) available through TCGA data portal (Archive: COADREAD.rnaseqv2__illuminaga_ promotes tumor progression and mesenchymal transition of human epithelial rnaseqv2__unc_edu__Level_3__RSEM_genes_normalized__data.data.txt). cells. Mol. Cell Biol. 21, 575–594 (2001). Expression levels were converted to log (base 2) after adding a small quantity (0.001) 13. Resar, L. M. The high-mobility group A1 gene: transforming inflammatory to account for zeros in the data. Boxplots, scatterplots and Spearman rank-based signals into cancer? Cancer Res. 70, 436–439 (2010). correlations were calculated using R statistical software suite using standard 14. Belton, A. et al. HMGA1 induces intestinal polyposis in transgenic mice and functions67. drives tumor progression and stem cell properties in colon cancer cells. PLoS ONE 7, e30034 (2012). 15. Schuldenfrei, A. et al. HMGA1 drives stem cell, inflammatory pathway, and cell Immunohistochemistry. Haematoxylin & eosin and IHC staining of organoid and cycle progression genes during lymphoid tumorigenesis. BMC Genomics 12, intestinal sections were performed after fixing organoid cultures in formalin at 4 °C 549 (2011). 14,16,68,69 before paraffin or frozen embedding . (See Supplementary Table 3 for list 16. Shah, S. N. et al. HMGA1 reprograms somatic cells into pluripotent stem cells of primary antibodies.) by inducing stem cell transcriptional networks. PLoS ONE 7, e48533 (2012). 17. Shah, S. N. & Resar, L. M. High mobility group A1 and cancer: potential 27, Time-lapsed confocal imaging and image analysis. Crypt cells were isolated biomarker and therapeutic target. Histol. Histopathol. 567–579 (2012). from WT and Hmga1 mice separately, resuspended in matrigel, and cultured on 18. Shah, S. N. et al. HMGA1: a master regulator of tumor progression in triple- glass-bottom 24-well plates41. Crypt cells were fixed either immediately (day 0) negative breast cancer cells. PLoS ONE 8, e63419 (2013). after seeding or following culture for 6 days42–45. Green fluorescence of Lgr5 þ 19. Johnson, K. R., Lehn, D. A., Elton, T. S., Barr, P. J. & Reeves, R. Complete (GFP þ ) cells was amplified by labelling anti-GFP antibody (00-106-215, Rockland murine cDNA sequence, genomic structure, and tissue expression of the high Immunochemicals Inc. Limerick, PA) followed by staining with Alexa Fluor 488 mobility group protein HMG-I(Y). J. Biol. Chem. 263, 18338–18342 (1988). anti-Goat secondary antibody (A-11055, ThermoFisher Scientific, Waltham, MA). 20. Friedmann, M., Holth, L. T., Zoghbi, H. Y. & Reeves, R. Organization, Entire organoids were visualized with Alexa Fluor 568 Phalloidin (A12380, inducible-expression and chromosome localization of the human HMG-I(Y) ThermoFisher Scientific) and Hoescht 33258 labelling (H3569, ThermoFisher nonhistone protein gene. Nucleic Acids Res. 21, 4259–4267 (1993). Scientific) using confocal imaging42–45. Images were collected using a Solamere 21. Pedulla, M. L., Treff, N. R., Resar, L. M. & Reeves, R. Sequence and analysis of Technology Group spinning disk confocal microscope with a Â 10 objective (Zeiss the murine Hmgiy (Hmga1) gene locus. Gene 271, 51–58 (2001). Microimaging). A combination of mManager and Piper (Stanford Photonics) 22. Thanos, D. & Maniatis, T. Virus induction of human IFN beta gene expression imaging software was used for imaging acquisition42–45. All organoids were imaged requires the assembly of an enhanceosome. Cell 83, 1091–1100 (1995). using identical excitation and detection settings per fluorescence channel. ImageJ 23. Falvo, J. V., Thanos, D. & Maniatis, T. Reversal of intrinsic DNA bends in the was used for post-acquisition image analysis and quantification of Lgr5 þ (green IFN beta gene enhancer by transcription factors and the architectural protein channel) Z-projected surface areas (mm2) per organoid. Briefly, 2 mm-step Z-stacks HMG I(Y). Cell 83, 1101–1111 (1995).

14 NATURE COMMUNICATIONS | 8:15008 | DOI: 10.1038/ncomms15008 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/ncomms15008 ARTICLE

24.Huth,J.R.et al. The solution structure of an HMG-I(Y)-DNA complex defines a 57. Akaboshi, S. et al. HMGA1 is induced by Wnt/beta-catenin pathway and new architectural minor groove binding motif. Nat. Struct. Biol. 4, 657–665 (1997). maintains cell proliferation in gastric cancer. Am.J.Pathol.175, 1675–1685 (2009). 25. Chiappetta, G. et al. High level expression of the HMGI (Y) gene during 58. Ullah, M. et al. Transdifferentiation of mesenchymal stem cells-derived embryonic development. Oncogene 13, 2439–2446 (1996). adipogenic-differentiated cells into osteogenic- or chondrogenic-differentiated 26. Ben-Porath, I. et al. An embryonic stem cell-like gene expression signature in cells proceeds via dedifferentiation and have a correlation with cell cycle poorly differentiated aggressive human tumors. Nat. Genet. 40, 499–507 (2008). arresting and driving genes. Differentiation 85, 78–90 (2013). 27. Zhou, G. et al. The pattern of gene expression in human CD34 þ stem/ 59. Buczacki, S. J. et al. Intestinal label-retaining cells are secretory precursors progenitor cells. Proc. Natl Acad. Sci. USA 98, 13966–13971 (2001). expressing Lgr5. Nature 495, 65–69 (2013). 28. Chou, B. K. et al. Efficient human iPS cell derivation by a non-integrating 60. Zeitels, L. R. et al. Tumor suppression by miR-26 overrides potential oncogenic plasmid from blood cells with unique epigenetic and gene expression activity in intestinal tumorigenesis. Genes Dev. 28, 2585–2590 (2014). signatures. Cell Res. 21, 518–529 (2011). 61. Wang, D. & DuBois, R. N. The role of anti-inflammatory drugs in colorectal 29. Munoz, J. et al. The Lgr5 intestinal stem cell signature: robust expression of cancer. Annu. Rev. Med 64, 131–144 (2013). proposed quiescent ‘ þ 4’ cell markers. EMBO J. 31, 3079–3091 (2012). 62. Feng, Y. et al. Sox9 induction, ectopic Paneth cells, and mitotic spindle axis 30. Wood, L. J. et al. HMG-I/Y, a new c-Myc target gene and potential oncogene. defects in mouse colon adenomatous epithelium arising from conditional Mol. Cell Biol. 20, 5490–5502 (2000). biallelic Apc inactivation. Am. J. Pathol. 183, 493–503 (2013). 31. Fusco, A. & Fedele, M. Roles of HMGA proteins in cancer. Nat. Rev. Cancer 7, 63. Kopp, J. L. et al. Identification of Sox9-dependent acinar-to-ductal 899–910 (2007). reprogramming as the principal mechanism for initiation of pancreatic ductal 32. Roy, S. et al. HMGA1 overexpression correlates with relapse in childhood B-lineage adenocarcinoma. Cancer Cell 22, 737–750 (2012). acute lymphoblastic leukemia. Leuk. Lymphoma 54, 2565–2567 (2013). 64. Lois, C., Hong, E. J., Pease, S., Brown, E. J. & Baltimore, D. Germline 33. Grade, M. et al. Gene expression profiling reveals a massive, aneuploidy- transmission and tissue-specific expression of transgenes delivered by lentiviral dependent transcriptional deregulation and distinct differences between lymph vectors. Science 295, 868–872 (2002). node-negative and lymph node-positive colon carcinomas. Cancer Res. 67, 65. Andersson-Rolf, A., Fink, J., Mustata, R. C. & Koo, B. K. A video protocol of retroviral 41–56 (2007). infection in primary intestinal organoid culture. J. Vis. Exp. 2014, e51765 (2014). 34. Hristov, A. C. et al. HMGA1 correlates with advanced tumor grade and decreased 66. Cartharius, K. et al. MatInspector and beyond: promoter analysis based on survival in pancreatic ductal adenocarcinoma. Mod. Pathol. 23, 98–104 (2010). transcription factor binding sites. Bioinformatics 21, 2933–2942 (2005). 35. Xu, Y. et al. The HMG-I oncogene causes highly penetrant, aggressive lymphoid 67. Hillion, J. et al. The high mobility group A1 (HMGA1) gene is highly malignancy in transgenic mice and is overexpressed in human leukemia. overexpressed in human uterine serous carcinomas and carcinosarcomas and Cancer Res. 64, 3371–3375 (2004). drives matrix metalloproteinase-2 (MMP-2) in a subset of tumors. Gynecol. 36. Tesfaye, A. et al. The high-mobility group A1 gene up-regulates cyclooxygenase Oncol. 141, 580–587 (2016). 2 expression in uterine tumorigenesis. Cancer Res. 67, 3998–4004 (2007). 68. Hillion, J. et al. The high-mobility group A1a/signal transducer and activator of 37. Dominici, M. et al. Transgenic mice with pancellular enhanced green transcription-3 axis: an achilles heel for hematopoietic malignancies? Cancer fluorescent protein expression in primitive hematopoietic cells and all blood cell Res. 68, 10121–10127 (2008). progeny. Genesis 42, 17–22 (2005). 69. van Es, J. H., de Geest, N., van de Born, M., Clevers, H. & Hassan, B. A. 38. Clevers, H. & Nusse, R. Wnt/beta-catenin signaling and disease. Cell 149, Intestinal stem cells lacking the Math1 tumour suppressor are refractory to 1192–1205 (2012). Notch inhibitors. Nat. Commun. 1, 18 (2010). 39. Kabiri, Z. et al. Stroma provides an intestinal stem cell niche in the absence of epithelial Wnts. Development 141, 2206–2215 (2014). Acknowledgements 40. Wu, C. et al. RSPO2-LGR5 signaling has tumour-suppressive activity in colorectal cancer. Nat. Commun. 5, 3149 (2014). We dedicate this work to the memory of our esteemed colleague and co-author, 41. Sato, T. & Clevers, H. Primary mouse small intestinal epithelial cell cultures. Dr David Huso. His generosity, kindness and intellect were immense and will continue to Methods Mol. Biol. 945, 319–328 (2013). inspire all who knew him. This work was supported by grants from the National Institute 42. Nguyen-Ngoc, K. V. et al. 3D culture assays of murine mammary branching of Health (CA2149550, CA164677, DK104344) and Maryland Stem Cell Research Fund morphogenesis and epithelial invasion. Methods Mol. Biol. 1189, 135–162 (2015). (2011-MSCRFE-0102, 2015-MSCRFE-1759). 43. Huebner, R. J., Lechler, T. & Ewald, A. J. Developmental stratification of the mammary epithelium occurs through symmetry-breaking vertical divisions of Author contributions apically positioned luminal cells. Development 141, 1085–1094 (2014). L.X. designed and performed most of the experiments, analysed the data and prepared 44. Ewald, A. J. Practical considerations for long-term time-lapse imaging of the manuscript; T.H., A.B., Y.-T.C., Q.G., X.Z., W.K. and D.L.H. performed experiments epithelial morphogenesis in three-dimensional organotypic cultures. Cold and analysed the data, L.C. analysed TCGA databases, S.S. and A.F. helped with organoid Spring Harb. Protoc. 2013, 100–117 (2013). cultures; D.G. and A.J.E. designed and performed experiments with confocal, single cell 45. Cheung, K. J. et al. Polyclonal breast cancer metastases arise from collective and organoid imaging, and analysed confocal microscopic data, L.M.S.R. conceived the dissemination of keratin 14-expressing tumor cell clusters. Proc. Natl Acad. Sci. study, supervised the project, edited and finalized the manuscript. USA 113, E854–E863 (2016). 46. Fuerer, C. & Nusse, R. Lentiviral vectors to probe and manipulate the Wnt signaling pathway. PLoS ONE 5, e9370 (2010). Additional information 47. Mori-Akiyama, Y. et al. SOX9 is required for the differentiation of paneth cells Supplementary Information accompanies this paper at http://www.nature.com/ in the intestinal epithelium. Gastroenterology 133, 539–546 (2007). naturecommunications 48. Formeister, E. J. et al. Distinct SOX9 levels differentially mark stem/progenitor populations and enteroendocrine cells of the small intestine epithelium. Am. J. Competing interests: The authors declare no competing financial interests. Physiol. Gastrointest. Liver Physiol. 296, G1108–G1118 (2009). 49. Bastide, P. et al. Sox9 regulates cell proliferation and is required for Paneth cell Reprints and permission information is available online at http://npg.nature.com/ differentiation in the intestinal epithelium. J. Cell. Biol. 178, 635–648 (2007). reprintsandpermissions/ 50. Roche, K. C. et al. SOX9 maintains reserve stem cells and preserves How to cite this article: Xian, L. et al. HMGA1 amplifies Wnt signalling and expands the radioresistance in mouse small intestine. Gastroenterology 149, 1553–1563.e10 intestinal stem cell compartment and Paneth cell niche. Nat. Commun. 8, 15008 (2015). doi: 10.1038/ncomms15008 (2017). 51. Gracz, A. D. et al. A high-throughput platform for stem cell niche co-cultures and downstream gene expression analysis. Nat. Cell. Biol. 17, 340–349 (2015). Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in 52. Di Cello, F. et al. HMGA2 participates in transformation in human lung cancer. published maps and institutional affiliations. Mol Cancer Res. 6, 743–750 (2008). 53. Hristov, A. C. et al. HMGA2 protein expression correlates with lymph node metastasis and increased tumor grade in pancreatic ductal adenocarcinoma. This work is licensed under a Creative Commons Attribution 4.0 Mod. Pathol. 22, 43–49 (2009). International License. The images or other third party material in this 54. Morishita, A. et al. HMGA2 is a driver of tumor metastasis. Cancer Res. 73, article are included in the article’s Creative Commons license, unless indicated otherwise 4289–4299 (2013). in the credit line; if the material is not included under the Creative Commons license, 55. Kugel, S. et al. SIRT6 suppresses pancreatic cancer through control of Lin28b. users will need to obtain permission from the license holder to reproduce the material. Cell 165, 1401–1415 (2016). To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/ 56. Bush, B. M., Brock, A. T., Deng, J. A., Nelson, R. A. & Sumter, T. F. The Wnt/beta-catenin/TCF-4 pathway upregulates HMGA1 expression in colon cancer. Cell Biochem. Funct. 31, 228–236 (2013). r The Author(s) 2017

NATURE COMMUNICATIONS | 8:15008 | DOI: 10.1038/ncomms15008 | www.nature.com/naturecommunications 15